| Age | Commit message (Collapse) | Author |
|---|
|
Needed by the CS checker.
|
|
|
|
We've supported indirect rendering pbuffers for a while, but not direct
rendering pbuffers. The way we do this is by creating a hidden pixmap
and wrap that in a GLX pbuffer. This only works when we have DRI2 on
the server, but if the server doesn't have DRI2, it won't expose configs
with pbuffer bits enabled.
|
|
When matching attributes using the 'mask' matching criteria, the spec
says that
"Only GLXFBConfigs for which the set bits of attribute include all
the bits that are set in the requested value are
considered. (Additional bits might be set in the attribute)."
The current test returns true if the two bit masks have bits in
common, specifically it matches even if the requested value has bits
set that are not set in the fbconfig attribute. For example, an
application asking for
GLX_DRAWABLE_TYPE, GLX_PIXMAP_BIT | GLX_PBUFFER_BIT,
as glxpbdemo does, will match fbconfigs that don't support pbuffer
rendering, as long as they support pixmap rendering.
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
|
|
Update the warranty disclaimer to use the more general "THE AUTHORS OR
COPYRIGHT HOLDERS". This is done manually on files created by me. Hope
that I do not miss anything.
|
|
|
|
|
|
|
|
|
|
We were looping over instructions but only looking at the 0th instruction's
opcode.
Fixes fd.o bug 27566.
|
|
|
|
More accurate results.
We should probably generalize these functions for more cases.
|
|
Note the tests don't test all the paths, in particular stride, and
ensuring the old value is preserved.
|
|
|
|
|
|
|
|
Depth-stencil manually written given that each one is very close to
be a special case.
u_format_zs.c's still untested.
|
|
|
|
And fix nv50_screen.c to compile against the updated header.
|
|
Should be an integer param, according to docs.
|
|
xorg st needs them.
|
|
Caused fun linker errors on r600g.
|
|
native.h is getting a little messy over time.
|
|
A pbuffer is an EGL resource. It does not need a native display to
create.
|
|
|
|
Commit 1454f20a991ddda35f1a2ffda953012078b407ba caused the regression.
When the vertex shader emitted both a texcoord and color we were grabbing
the wrong vertex attributes. Fix the draw_quad() code to put texcoords
in slot[1] and color in slot[2]. That's a bit cleaner than changing
the vertex shader code.
Tested with progs/tests/zreaddraw.c
|
|
|
|
|
|
Make sure we have up to date buffers before we start looking at
the tiling bits to determine how to clear.
|
|
Untested.
|
|
|
|
|
|
|
|
|
|
Makes it easier to copy and paste.
|
|
Make sure we have up to date buffers before we start looking at
the tiling bits to determine how to clear.
|
|
we were only using the jited function in the linear case, now drawelts
correctly uses the same path. it results in a significant gain in
real world apps (openarena went from 23fps to 29fps)
|
|
|
|
successful."
Otherwise the library will be unloaded and function pointers become invalid.
This reverts commit bc2bc0306e4dd8c56bd66a8aabf2433f6689653d.
|
|
|
|
|
|
|
|
Conflicts:
Makefile
configs/default
src/mesa/main/version.h
|
|
When the format is __DRI_TEXTURE_FORMAT_RGB, the texture should be
treated as if there is no alpha channel.
|
|
piglit has a test called fbo-drawbuffers, this fails for me on r300g,
and fixing the texenv program to use the DATA outputs fixes it.
Signed-off-by: Dave Airlie <airlied@redhat.com>
|
|
our code resets pipe_vertex_buffer's with different offsets when rendering
vbo, meaning that we kept creating insane number of shaders even for simple
apps e.g. geartrain had 54 shaders and it was taking almost 27 seconds just to
compile them. this patch passes pipe_vertex_buffer's to the jit function and lets
it to the stride/buffer_offset computation at run time. the slowdown at runtime
is largely unnoticable but the we go from 54 shaders to 3, and from 27 seconds to less
than 1.
|
|
|
|
|
|
|
|
Unnecessary, and doesn't even guarantee size.
|
